Productivity Blueprint
Persistent Project Memory with Daily AI Sync
Keep project context persistent across AI conversations without losing details. A daily sync pulls updates from meetings, Slack, and tickets into organized living docs that automatically load at sessi
What problem does this solve?
Managing many ongoing projects through separate AI chat sessions puts all the context inside the chat thread, where it is lost between conversations and eventually overflows the model's context window. Reconstructing where each project stands from scattered meeting recordings, Slack messages, tickets, and documents is slow and easy to get wrong.
How does it work?
1. Create one folder per project and keep a few living documents in each (an overview, running notes, a decisions log, and a task list) as the durable source of truth instead of the chat. 2. Connect that folder to Claude and have it read the living docs at the start of any session, so a fresh conversation loads full context without depending on past chat history. 3. Schedule a daily task in Claude that pulls new activity from meeting recordings, Slack channels, Jira, Confluence, and Google Docs and files it into each project's notes, decisions, and tasks. Have it break a single multi-topic call into its separate topics and route only the relevant slice to each project. 4. Have Claude build an Obsidian knowledge vault inside each project folder: an interlinked graph of the project's people, decisions, features, systems, and open issues with a Home entry point that an agent can traverse. 5. Add a CLAUDE.md instructions file at the folder root so every new session automatically reads the right project's knowledge graph first and stays focused. 6. Schedule a weekly task that reconciles each knowledge graph against the latest living docs and validates that every link resolves, so the graph never drifts or breaks. 7. Start each work session fresh, connect the folder, name the project, and let the loaded context carry it forward.
What's the biggest win?
Project knowledge stays current automatically and lives in durable files rather than in any single conversation. New sessions begin with full context, long chats no longer overflow or lose history, and each morning brings a short per-project briefing of what changed and what to prioritize next.
What should I know technically?
Treat the living docs as the single source of truth and each chat as disposable, writing new information to files rather than the thread. In the daily task, use the most recent date in each project's notes as a watermark so pulls are incremental and idempotent, and always sweep at least the previous day's calls so nothing is missed after weekends. Store each project's source list (Slack channels, Jira keys, document links) in its overview file so the task self-discovers what to pull, and scope connector queries tightly, for example limiting Jira to your own project plus any cross-team epics a project explicitly cites. Generate each knowledge graph with a script that enforces uniform frontmatter and validates that every link resolves with no orphans. Keep the daily task's edits to the graph light and append-only, and run the heavier reconcile weekly so an unattended run cannot break links.
What are the constraints?
Freshness is daily rather than real-time, and scheduled runs depend on the app being open at the scheduled time. Coverage is limited to the sources each project lists and the connectors that are authorized, so anything outside them is not captured. The knowledge graph can lag the source documents between refreshes, and files on disk can still be moved or deleted, so a backup or version-control layer is recommended.
About This Blueprint
- Industry
- Information Technology